Klaus-Dieter Kuhnert

Robust adaptive control of nonholonomic mobile robot with parameter and nonparameter uncertainties

This paper considers the tracking-control problem of a nonholonomic wheeled mobile robot with both parameter and nonparameter uncertainties. A robust adaptive controller is proposed with the aid of the adaptive backstepping technique and the learning ability of neural networks. The proposed controller guarantees that the tracking error converges to a small hall containing the origin. The halls radius can be adjusted by control parameters. The proposed controller is successfully implemented in our simulator.

Fusion of Stereo-Camera and PMD-Camera Data for Real-Time Suited Precise 3D Environment Reconstruction

3D environment reconstruction is a basic task, delivering the data for mapping, localization and navigation in mobile robotics. We present a new technique that combines a stereo-camera system with a PMD-camera. Both systems generate distance images of the environment but with different characteristics. It is shown that each system compensates effectively for the deficiencies of the other one. The combined system is real-time suited. Experimental data of an indoor scene including the calibration procedure are reported

A Vision System for Real Time Road and Object Recognition for Vehicle Guidance

One crucial component of a control system for autonomous vehicle guidance is real time image analysis. This system part is burdened by the maximum flow of information. To overcome the high demands in computation power a combination of knowledge based scene analysis and special hardware has been developed. The use of knowledge based image analysis supports real time processing not by schematically evaluating all parts of the image, but only evaluating those which contain relevant information. This is due to the fact that in many practical problems the relevant information is very unevenly distributed over the image. Preknowledge of the problem or the aim of the mission and expectations or predictions about the scene sustantially reduce the amount of information to be processed. The operations during such an analysis may be divided into two classes - simple processes, e.g. filters, correlation, contour processing and simple search strategies - complex search and control strategy This classification supplied the concept for a special hardware. The complex tasks are performed by a universal processor 80286 while the remaining tasks are executed by a special coprocessor (including image memory). This combination permits the use of filter masks with a arbitrary geometry together with a powerful search strategy. A number of these basic modules may be configured into a multiprocessor system. The universal processor is programmed in a high level language. To support the coprocessor a set of software tools has been built. They permit interactive graphical manipulation of filtermasks, generation of simple search strategies and non real time simulation. Also the real data structures that control the function of the coprocessor are generated by this software package. The system is used within our autonomous vehicle project. One set of algorithms tracks the border lines of the road even if they are broken or disturbed by dirt. Also shadows of bridges crossing the road are tolerated. Another algorithm tracks prominent points on other objects (e.g. vehicles) to collect possible candidates of obstacles during the real time run. A complete image analysis for the relevant features is performed in one video cycle (16.6 ms).

A lane change detection approach using feature ranking with maximized predictive power

Risk estimation for the current traffic situation is crucial for safe autonomous driving systems. One part of the uncertainty in risk estimation is the behavior of the surrounding traffic participants. In this paper we focus on highway scenarios, where possible behaviors consist of a change in acceleration and lane change maneuvers. We present a novel approach for the recognition of lane change intentions of traffic participants. Our novel approach is an extension of the Naïve Bayesian approach and results in a generative model. It builds on the relations to the directly surrounding vehicles and to the static traffic environment. We obtain the conditional probabilities of all relevant features using Gaussian mixtures with a flexible number of components. We systematically reduce the number of features by selecting the most powerful ones. Furthermore we investigate the predictive power of each feature with respect to the time before a lane change event. In a large scale experiment on real world data with over 160.781 samples collected on a test drive of 1100km we trained and validated our intention prediction model and achieved a significant improvement in the recognition performance of lane change intentions compared to current state of the art methods.

Vision-based autonomous road vehicles

Autonomous road vehicles, guided by computer vision systems, are a topic of research in numerous places in the world. Experimental vehicles have already been driven automatically on various types of roads. Some of these vehicles are briefly introduced, and one is described in more detail. Its dynamic vision system has enabled it to reach speeds of about 100 km/h on highways and 50 km/h on secondary roads.

Structure overview of vegetation detection. A novel approach for efficient vegetation detection using an active lighting system

Fully autonomous navigation has been widely investigated for several decade of years; however, a safe and reliable navigation is still a daunting challenge in terrains containing vegetation. To improve the mobility capability of recent autonomous navigation systems, an additional vegetation detection function has been proposed. Since many proposals of generating vegetation classifier as well as suggestions of using different sensors to implement the function exist, a structured overview is required for vegetation detection in the context of outdoor navigation. Therefore, this paper studies and compares the accuracy and efficiency of existing vegetation detection approaches in a structured way. Furthermore, a new vision system set-up which combines CMOS sensor and Photo Mixer Device sensor with a near-infrared lighting system is also introduced to simultaneously provide depth, near-infrared and color images at high frame rate. Those near-infrared and color information are then used to compute vegetation index or train vegetation classifier to completely realize a real-time robust vegetation detection system. In this paper, a modification of the normalized difference vegetation index is devised, which is then defined as the new standard form of vegetation index for such vision system integrated with an additional lighting system. Finally, we will show the out-performance of the proposed approach in comparison with more conventional ones.

A novel approach for a double-check of passable vegetation detection in autonomous ground vehicles

The paper introduces an active way to detect vegetation which is at front of the vehicle in order to give a better decision-making in navigation. Blowing devices are to be used for creating strong wind to effect vegetation. Motion compensation and motion detection techniques are applied to detect foreground objects which are presumably judged as vegetation. The approach enables a double-check process for vegetation detection which was done by a multi-spectral approach, but more emphasizing on the purpose of passable vegetation detection. In all real world experiments we carried out, our approach yields a detection accuracy of over 98%. We furthermore illustrate how the active way can improve the autonomous navigation capabilities of autonomous ground vehicles.

Off-road Robotics—An Overview

This article gives an overview of the current state of research in the field of off-road robotics. It focuses on techniques used in the areas of perception, environment representation, as well as navigation, and introduces different types of robot control systems. A presentation of different applications is given along with an outlook on future developments.

Fusing dynamic vision and landmark navigation for autonomous driving

Basic concepts for navigation and orientation of autonomous mobile systems are the main subject of this paper. Landmark navigation helps to integrate singular observations into a global world model. In a vision-guided vehicle, observations may stem from dynamic vision allowing the control of movements by a video camera in real-time. The paper describes the combination of both approaches to an homogeneous system which is implemented in the experimental vehicle ATHENE. Considerations on the appropriate knowledge representation, the data structures, and aspects relevant to real-time implementation of the various maps or control systems are presented.<<ETX>>

When will it change the lane? A probabilistic regression approach for rarely occurring events

Understanding traffic situations in dynamic traffic environments is an essential requirement for autonomous driving. The prediction of the current traffic scene into the future is one of the main problems in this context. In this publication we focus on highway scenarios, where the maneuver space for traffic participants is limited to a small number of possible behavior classes. Even though there are many publications in the field of maneuver prediction, most of them set the focus on the classification problem, whether a certain maneuver is executed or not. We extend approaches which solve the classification problem of lane-change behavior by introducing the novel aspect of estimating a continuous distribution of possible trajectories. Our novel approach uses the probabilities which are assigned by a Random Decision Forest to each of the maneuvers lane following, lane change left and lane change right. Using measured data of a vehicle and the knowledge of the typical lateral movement of vehicles over time taken from realworlddata, we derive a Gaussian Mixture Regression method. For the final result we combine the predicted probability density functions of the regression method and the computed maneuver probabilities using a Mixture of Experts approach. In a large scale experiment on real world data collected on multiple test drives we trained and validated our prediction model and show the gained high prediction accuracy of the proposed method.